It’s with surprise that I read a post on The Register and another on TUAW regarding the iPhone 3G‘s speed, or lack thereof. The way 3G networks operate has some fundamental differences compared to GPRS or EDGE, the so-called 2G or 2.5G networks. A word of caution before you proceed, if you are a tech purist – I am explaining this in the simplest terms I can find, so don’t be offended by over-simplification.
“Scotty, we’ll need everything you’ve got.” 
Let me bore you with a primer on the “old” technology. In the very old days, packet data didn’t exist on GSM cellular networks. Each GSM frequency (also known as physical channel) would be split into eight full-rate (remeber the old EFR hacks?) timeslots, or sixteen half-rate. Voice was digitized using a vocoder, and transmitted to the cell tower using one of the timeslots. The tower then passed it along the switching network, to be terminated at another GSM terminal, or sent out to the PSTN where it could also be converted back to analog for further relay to landline phones. SMS would be sent over the control channel, thus not taking up timeslots for voice calls.
When data was first introduced to GSM, it was in the form of circuit-switched calls, the equivalent to the analog modems used in the early internet days. Since this was the equivalent of dialing a landline modem, the network could only provide one timeslot, and a brute rate of 9.6kbps (yeah, imagine the iPhone running on that!).
Then came GPRS, which brought packet-data to the phone, allowing the adjoining of several timeslots into one single packet-switched call, theoretically giving rates of 56 to 114kbps, which in reality could be as low as 9.6kbps, as cell operators would prioritize voice calls over data. The same occupation time of a timeslot on voice could earn the operator up to 10 times more than the same slot used for data. Thus, in the days when cell density was very low, and voice capability scarce, it was not uncommon to find that a tower would have only one timeslot reserved for GPRS and the rest for voice, sending you back to switched-call speeds.
I am the EDGE! 
When new modulation schemes and tighter timing capabilities made it possible, GPRS was upgraded to EDGE, which is an acronym for Enhanced Data rates for GSM Evolution. The theoretical maximum speed a packet-switched connection could now achieve was raised to 236.8kbps, with an upload speed of 59.2kbps. The upload and download speeds can be balanced to give a more symmetric connection, but this is rarely used.
Remember that we are still talking about chopping up a finite resource, a radio frequency channel, into smaller chunks, each of which can now carry data at increasing speeds, and also be aggregated to boost the overall transfer rate.
Slow? Twelve minutes? Hannibal, I had to say hello! 
One day in some lab, a fool invented CDMA. Well, not a fool, a very clever individual – just that the use of CDMA has been taken out of its initial expectations.
CDMA stands for Code Division Multiple Access. Contrary to GSM which uses TDMA, or Time Division Multiple Access, and divides the spectrum into timed slots for use by one mobile station at a time, CDMA allows ALL stations to talk to to the tower at once, with each given an individual code that is embedded in all transmissions. Then, using this code, the tower and terminal can tell who’s traffic it is amidst all the noise.
OK, it’s hard to imagine – let’s try a practical example. Imagine a room full of people, all talking at once to a host. The host gives each member of the public a phrase book in a different language, which they must use to translate what they want to say to the host. The host can then attempt to figure out who is saying what by figuring out the language of each utterance, and translating it back to the original. Sounds complex? You bet! The size of the room was also increased, whereas GSM used 200kHz of spectrum for its eight timeslots, 3G/UMTS using CDMA uses 5MHz, 25 times more spectrum, to achieve a theoretical maximum 14Mbps throughput. In practice, 3G HSDPA has only been able to achieve 7.2Mbps download so far, with HSUPA trying to improve the upload speed.
Before going further, if we compare the transfer speeds of each technology, EDGE can receive 1.18 kb/kHz, compared to 1.44 kb/kHz for HSDPA. Nothing to write home about, considering the huge costs this technology is having, I would say!
Stop it already, tell me what’s wrong with my iPhone!
What is basically wrong is that you have a beautiful screen on a beautiful mobile phone, capable of rendering full-size web pages…but you are sharing your internet connection with all your neighbors. In essence, 3G is allowing everyone to take a piece of the spectrum and use it to send and receive voice and data. This means a single 3G tower can handle a huge ammount of voice traffic, as it uses fewer resources, but when data is also added, it can be strongly degraded. I would venture that operators still prioritize voice over data on their 3G networks.
The key issue to remember is that the download rate is “per tower”, not per user. So, if two users using HSDPA are on the same tower, they will each get a maximum throughput of 3.6Mbps. Divide even further, and the more users you have the worse experience everyone gets. Further add tons of voice calls into the mix, and you have what The Register and TUAW describe in their posts – frustration. The blame may be at the iPhone’s radio, but from what I have been experiencing in Spain, Telefonica’s GPRS network provides almost as good speed as 3G in the very busy Barcelona. Why? Because my GPRS connection is getting better, as more voice traffic is dumped on the 3G networks, and I have more slots to myself!
- Scotty in Star Trek, first series.
- Aeon Flux in, well, Aeon Flux.
- Face in The A-Team.